We developed and tested an algorithm to automatically classify twenty runners as novice or experienced based on their technique. Linear accelerations and angular velocities collected from six common wearable sensor locations were used to train support vector machine classifiers. The model using input data from all six sensors achieved a classification accuracy of 98.5% (10 km/h running). The classification performance of models based on single sensor data showed a 56.3-94.5% accuracy range, with sensors from the upper body giving the best results. Comparisons of kinematic variables between the two populations confirmed significant differences in upper body biomechanics throughout the stride, thus showing applied potential when aiming to compare novice runner’s technique with movement patterns more akin to those with greater experience.
New Investigator Award
Carter, Josh A. Mr; R Rivadulla, Adrian; and Preatoni, Ezio
"SUPPORT VECTOR MACHINES CAN CLASSIFY RUNNER’S ABILITY USING WEARABLE SENSOR DATA FROM A VARIETY OF ANATOMICAL LOCATIONS,"
ISBS Proceedings Archive: Vol. 39
, Article 72.
Available at: https://commons.nmu.edu/isbs/vol39/iss1/72